wmmmn}mi¥^7%^^
u^ r
<^S^CM&^^
o^y
M.I.T.
LIBRARIES
-
DEWEY
f.
oewvev
D28
M414
^3
ALFRED
P.
WORKING PAPER
SLOAN SCHOOL OF MANAGEMENT
The Design Spaces
of
Audio Menus and Forms
Paul Resnick*
Robert A. Virzi**
MIT
MIT Center
Sloan School Working Paper #3633-93
for
Coordination Science Technical Report #159
November, 1993
MASSACHUSETTS
INSTITUTE OF TECHNOLOGY
50 MEMORIAL DRIVE
CAMBRIDGE, MASSACHUSETTS 02139
The Design Spaces
of
Audio Menus and Forms
Paul Resnick*
Robert A. Virzi**
MIT
MIT Center
Sloan School Working Paper #3633-93
for Coordination Science Technical Report #159
November, 1993
Abstract
Menus and forms
are important dialogue structures in telephone-based interactive
voice response and other audio applications. There is a surprising lack of diversity,
however, in the interaction styles they employ. This article presents design spaces
for audio menu and form styles. The key idea is to break recordings and actions into
parts. The methods of recombining the parts are the dimensions of the design
spaces. Twelve alternative menu styles and five form styles illustrate some of the
recombination possibilities. Choices on each of the design dimensions affect user
interactions, the sound and feel, in predictable ways. The best style will depend on
the experience levels of an application's users.
*MIT Room E53-325
Cambridge, MA 02139
presnick@mit.edu
**GTE Laboratories Incorporated
Waltham, MA 02251
rvirzi@gte.com
M.I.T.
JAN
LIBRARIES
3
1994
RECi-IiVED
1.
Introduction
Digital storage and processing of audio have
applications. There
opened new
possibilities for
already a large and growing market for telephone-based voice mail
is
and interactive voice response services. With the advent of personal
the integration of audio into desktop computing, speech
for
eyes-busy
speech-based
applications,
for
is
digital assistants
and
also likely to gain importance
personal communications,
and for records of
conversations.
Compared
provides no short term
memory
aids [Halstead-Nussloch 1989;
readers can read faster than they can listen.
that
can aid
increasing
speech output
to visual presentation of information,
listeners.
The
first is to
speedup while
Some
still
Schmandt
In press].
Good
technologies are available, however,
play speech back faster than
number of voice mail systems
pitch distortion.
Two
slow, serial, and
is
it
was recorded. An
offer accelerated playback, usually without
digital signal processing
mechanisms allow
retaining intelligibility [Arons 1992; Kato and
The second, and more important technology,
is
forward a conventional audio cassette tape, but
it
random
a factor of
Hosoya
access.
It
two
1993].
takes time to fast
takes virtually no time to
jump
to a
different part of a digitally stored recording. Meaningful subdivision of recordings,
together with user control
overjumps between those
parts of a recording entirely. Elsewhere,
parts, allows listeners to skip
we have described
as skip
some
and scan those audio
interfaces that allow users to scan a recording by skipping frequently [Resnick and Virzi
1992; Virzi,
et al.
].
[Arons 1993] explores playback controls
that affect both
speed of
playback and skips between segments.
To
exploit
recordings.
random
access, a designer needs to identify meaningful segments in
One source of segmentation
is
pre-defined structure, such as the separate
entr>'
blanks
[Degen.
in a
form.
also indicate
Gould and Boies 1983; Stifelman,
1992;
et al.
The person recording can
segments
1993]. In
et al.
some
infer
as turn taking
between speakers [Hindus and Schmandt 1993].
This article describes what user control, through random access, can do for two
audio dialogue structures, menus and forms.
options from a pre-defined
information. Audio
set.
A
cases, a
from acoustic properties of a recording, such
computer can
after the fact
segment boundaries
A menu
common
allows selection of one or more
form allows entry of a collection of related pieces of
menus and forms present options and
instructions through spoken
voice and allow user input either through buttons or speech.
Many menu and form
styles are possible.
applications. There
a surprising lack of diversity,
We
employed.
human
is
The optimal choice
however,
is
in the styles
between
commonly
system builders, the fundamental
break dialogue components into smaller pieces, then reconstruct them
to
We
novel ways.
likely to vary
organize the design space to help system builders, interface designers, and
factors researchers explore the possibilities. For
strategy
is
show
that
many
styles
can be constructed from the same primitive
in
parts,
by varying a few features of how they are glued together. For interface designers, we
discuss
and
how
feel.
variations in those construction features affect user interaction, the
The preferred sound and
feel will differ
among
sound
users depending on their
experience with an interaction style and a particular application.
All our examples assume interaction over a telephone.
The
analysis applies, however, to
any audio presentation of menus and forms, over the phone or with some other device
such as a Personal Digital Assistant. As
even relevant
line
LCDs,
output.
to limited
we
will argue in the conclusion, the analysis
is
bandwidth visual output devices, such as 20 character by two-
since such devices have the
same temporal presentation
constraint as audio
The
critical factor that distinguishes
forms
our analysis from analyses of most visual menus and
the temporal presentation of information.
is
that users consider the options
one
at
Some
a time [Lee and
analyses of visual
menus assume
MacGregor 1985; Paap and Roske-
Hofstrand 1986], while others assume a more flexible process [Card 1982; Landauer and
Nachbar 1985], leading
menu
hierarchies [Kiger 1984; Miller 1981].
menu: a user
shifts attention
menu item
keypad input
.
at a
1986; Pierce, et
time; users controlled
some audio menu
styles give users control over
Our examples
al.
artificially restricted to
when
al.
to
1992].
move
temporal
The screen displayed
to the next
item with
The models of human search processes developed from those experiments
will likely apply to
symbol
this
between items by shifting eye gaze.
presentation of items [MacGregor, et
9) or
[Norman 1991] summarizes much of
experiments explored visual menu styles that were
only one
depth of
of these analyses, however, assume simultaneous presentation of the entire
literature. All
Two
to different conclusions about the optimal breadth or
all
(*
when
#).
all
for input, with the buttons referred to
Keys could also be labeled by
enter a letter sequence or
because not
audio
menu
to hear the next item.
assume touch-tones
and
styles but not others
letters,
word [Davis 1991; Detweiler,
by number (0-
however, so that users could
et al.
1990; Fast and Ballantine
1988; Maries 1990] to initiate actions. Buttons on a hand-held device would lead to
similar interactions.
where
it
We
briefly
mention input by speech recognition
in those situations
could offer significant advantages over buttons.
Sections 2 and 3 elaborate the system and user perspectives on the design of audio
dialogues, drawing on
to describe
menu
styles to illustrate. Sections
and analyze design spaces for menus and forms.
perspective summarizes existing styles and suggests
suggests
4 and 5 apply these perspectives
when
particular styles will be
most appropriate.
new
In both cases, the
ones.
system
The user perspective
2.
Overview: the System Perspective
The system perspective follows
mechanism
there are
into
its
a traditional
constituent parts and recombine the parts in novel ways. In this case,
two kinds of constituent
parts.
The
actions such as marking a particular item in a
blank
1}
in a
engineering approach. Divide a dialogue
first
menu
are voice recordings.
The second
are
or adding a value to a particular entry
form. Four design dimensions govern the recombination of these parts:
Action combinations: Which component actions and combinations can users
initiate?
2) Action Distribution:
From which
User Inaction: What
3)
effect
recordings are actions available?
does user inaction (a timeout) have from each of the
recordings?
4)
User Initiated Movement: What transitions between
the recordings can users
initiate?
We
introduce these ideas by applying them to the most popular implementation of audio
menus. Section 4
will recursively subdivide the recording for
will generate alternative
form
styles through the
each menu item. Section 5
same technique of subdivision and
recombination.
The predominant implementation of audio menus
describes
all
the options sequentially.
Any
menu.
division of the selection actions.
We
first
is
as a single recording that
time during playback of the recording, a caller
can press a number associated with an option
interaction with such a
currently
to select
it.
Example
1
presents a sample
consider division of the recording and then
;
Welcome
to the
ABC
Bank's bank-by-phone.
For account balances, press
To
money between
transfer
For mortgage
To open
This
a
accounts, press 2;
rates, press 3.
new
menu
1
will
account, press
now
repeat.
4.
^
Make your
selection at any time.
[presses 4]
1: A diagram of the standard menu style
implemented with separate recordings for the
header, menu items, and footer. Boxes indicate
recordings. When the system finishes playback
Figure
Example 1: the standard menu style.
System prompts appear in plain text,
while user actions are bracketed. The
symbol,
caret
^,
inserted
the
the key press
in
prompt, indicates when
is made. That is, the caller presses 4
before hearing that the menu will
repeat.
The recording can be divided
descriptions of each of the
may omit
the
into six parts: (1) an introductory header, (2)
items, and (6) a concluding footer.
Some menu
the header or the footer. Other divisions are possible, but this one
particularly natural.
actions,
menu
of the recording in the current box. it begins
playing the box just below it. Arrows inherit
from the outside in. Thus, the selection actions
are available from the header box. the item
boxes, and the footer box, even though they are
shown only once for the enclosing box.
one
that
Each selection action can also be divided
marks
a preference for an option
and one
into
-
(5)
styles
seems
two component
that terminates interaction with
menu.
Appropriate choices on the four design dimensions can duplicate the interaction style of
Example
1.
First,
keep the selection action as a composite of marking an option and
terminating interaction with the menu. Second,
from
all
make
all
the selection actions available
the recordings. Third, automatically transition to the next part
when
users are
passive. Fourth, do not allow any explicit user transitions between the parts. Figure
summarizes these choices
Consider
how
in a
notation that
a user interaction
begins by playing the
first
we
use throughout the paper.
would proceed, given those design choices. The system
part, the
header. If the user does nothing, the system
automatically transitions to playing the description of the
first
and so on. At any time, the user can press a numbered button
Moreover, there
is
1
option, then the second,
to select
any of the options.
nothing else the user can do but wait or select. There
example, to jump from the header
to the description
of the
Dividing a recording would not be very interesting
if
it
is
no way,
ror
last option.
only led to duplication of the
original interaction style. Other choices on the design dimensions, however, lead to
alternative interaction styles. Consider
First,
what
menu
style
if
possibilities.
separate marking and termination actions were provided?
combined these
a user could
some other
mark
.several
into a single select operation. If they
The standard
were separate, however,
items before terminating, thus selecting several items from the
menu.
Second,
how
style, all
of the parts inherited
are the selection actions distributed
all
among
the parts? In the standard
menu
of the selection actions. Another possibility, however,
to distribute the actions positionally, so that a
listening to the recording that describes
it.
A
menu item can only be
is
selected while
single button, then, could be used to select
the current item rather than providing separate selection buttons for each of the items.
Third, what effect does user inaction have? In standard menus, inaction caused an
automatic transition to the next
of the current recording.
A
part.
Another possibility
third possibility
explicitly reject each option; doing nothing
is
is
for inaction to cause repetition
for inaction to cause selection. Users
would cause
would
the current item to be selected.
Fourth, what transitions can users initiate between the parts? Figure
explicit user navigation
that
among
advances from one part
description of one
immediately
start
the parts.
to the
menu item by
Another possibility
next.
is
1
did not allow any
to provide a
"next" button
For example, a user could interrupt the
pressing the "next" button.
The system would
playing back the next item. Other explicit transitions are also possible,
such as jumping ahead by several options, or restarting the menu.
2.1
An
Alternative Style
Consider an alternative menu
we
caJl
it
and
its
style,
shown
variants 2-button styles.
in
One
Example
2.
Since
current item, and
it
it
uses only two buttons,
button advances to the next item.
button selects the current item. In terms of the design choices,
select button, but
it
it
still
The other
uses a composite
distributes the actions positionally, timeouts cause repetition of the
provides an explicit "next" button to transition to the next item.
Figure 2 summarizes the interaction
style.
Welcome
XYZBank's bank-by-phone. To
to
the first option, press 3.
hear
'^
[presses 3j
Account balances. To
select this option, press
^
For the next option, press
1.
3.
[presses 3 J
Transfers between
option, press
1.
accounts.
rates.
3.
interrupting
To
'^
[presses
a
new
select this
3.
3.
prompt j
select this option, press
For the next option, press
Open
To
For the next option, press
[presses
Mongage
'^
interrupting
account, ^
I.
3.
To
prompt again!
select this option, press
1.
[presses
Example
2:
1,
A
interrupting
prompt again]
sample 2-button menu
dialogue.
The 2-button menu style. When
no adjacent box, it replays the
contents of the current box. That is, the
current item will keep replaying until the
Figure
there
2:
is
user presses a button.
2.2
A Simple Subspace
Not
all
3
combinations of design choices yield plausible menu
summarizes
a subspace defined by
some of
styles, but
the possible choices.
many
do. Figure
The rows
indicate
three choices for transitions from one part to the next: automatic (timeout) transitions,
explicit user actions, or both.
selection actions
among
The columns indicate
the
the parts: absolute or positional.
composite selection action, so
that users
two choices
The
for distributing
entire subspace
can select only one item from a menu.
assumes a
The two
styles
transitions
above
fill
two of the
Absolute selection together with automatic
cells.
between parts defines the standard menu
style. Positional selection
along with
explicit transitions defines the 2-button style.
The other four
cells
In press] uses only
mix and match these
one button. The
features.
The temporal menu
listener waits while the
computer
style
[Schmandt
recites the options
and presses the select button upon hearing the desired option. The temporal with skips
style adds an explicit skip
ahead button
temporal menus, so that a listener can either
to
wait through the recitation of the options, or press a button to skip through them. The
standard with skips
style
works
just like standard
menus, but provides
in
addition a
button, say #, that a listener can press any time to skip ahead to the next option.
stepped numeric style removes the automatic advance, so that a listener can only
to the next option
move on
by pressing the skip key.
How
to
select
Numeric
Absolute
u
c
R3
>
The
Positional
Timeouts
Standard
Temporal
Skip key
Stepped
2-button
<s£
Temporal
<5:!
o
numeric
I
9
o
X
Timeouts
and
Standard
with
with
skips
skips
^
Skip key
Figure
3:
The subspace defined by
advance from one item
the
dimensions of
to the next.
10
how
users select items and
how
they
2.3 Dividing the Selection Action
Any
of the srvies
:e.-r_ - _
st>"!e is
-
:
_
-
:
a var.ir.:
±e
in
-
^
-
-
subsp^ce ibove could be
;
:r. „".;
- ir.
a composite se
!-?_:: ^n st>!e.
.
rr.ri::":ed [o
"
. :
Example
3
-cuon. The positional multi-selector
shows
a
sample interaction.
5
-77- _..
[Beepi Traisfers --
.:;;;.
--
rt--
-
-
use serarate markina and
;-.
-
3- If
-
vou're d -e
•
3. '^If
you'redooe„.
ess 3. If
Ej.i~.r.e
2.4
re r:s:::onal multi-selector st\le.
.-
Summary
of
The beep
indicates that an option
defines a space o: design choices.
Tne design dimensions come
:'::
-
in
novel ways. Four dimensions of choice govern the recombination:
;::::-.; are
-:
'
3
::--ework of subdividing recordings and actions and recombining them
.
combmed
into
:
:
-;•=-=-:
how
primitive
composite ones; whether actions are distributed posiiionally or
_-^:..:: -r--.inism
:--
is
System Perspective
The sys:e~ re-rec::ve
_- _' _
voure done
is
utilized: the effect
of timeouts: and what
commands
are
betv^een recordings.
Over^iev;: User Perspectives
From
-•-?-'
\'
.
:-~-
:
e-e
perspective,
s
•
:---re ""e
some system
- j-.d
and
feel
level choices
make
little
difference while
of an interface. In addition, prompt wording
11
which
stv'les.
are immareria- :::-. :-=
user perspective.
We
discxiss
how
on user
thai relate
syste—
:r
.
;-
interactions.
Experience Levels
3.1
All users are not alike.
They
':.-
-
:
how
to
execute them. For example, users
not need to hear a prompt that says
skills that
makes user control so
;:_
::t-t'-
Moreover, they require different infonr-:.
how
::
: r.
:.;.; .r.z-
.;
who
familiarit>'
to lake.
^re;:
selects
to tell
.isers"
and which moves on
them how
how
Tee second
—:
~:_t
interests -t.z
_-._:
:.:--.
—zz-
variable
who
are
b:m
at ail "because
:?
is
.-
^r-.t -s:
-
to initiate them.
12
;
-_:::n
rr.-.ris
decide v.--:
_:
:-
~i\ be aware of
?
—
^
-e
kr.cv-.
ail
i-; :: :ake wr.r
rem what acnons
me
_rr::rr.a:e.
the op'ior.s
kr.o\^ vs-w.- iceys are assc-,;..^:;^ - /.-
ac::
users'
users' famiiiarir)- with
me:r.__-.-m _-^ ;:-:;-: expens
±ey
.t is
have decided
to initiaie actions they
applicj::- :-
of them. Such content expens know which
how
.
to the next item. S_:r. m::r.^-.-r. exre.-_s r.eez
menu, even liough ihey may not
hear any voice recordings
_-_:.;_-
familiar with the 2-bur:cr.
to execute actions.
descriptions of them. Users
.-
desirable.
For example, frequent users of a voice mail
mam
to select zr :r::?- ::o
ir_:,:-
:
contents of the dialogue, the information necessar.
the
.
:w now
al.'-eaiy
with the mechanisms o: ihe dialogue,
For example,
-
^:
^
Figure 4 shows a graph of user tvpes. -t:.?.zi by r»c
on
:
users" experience levels
and then present general principles
st>'les to their effects
-•;- rer-re.:.
-;.
may
:
_
>:me
-z^.r.z
-:: -ee^ :o
are avi:;i.ble
ana
i
High
Content Expert/
Expert
Mechanism Novice
Content
Low
Hich
Familiarity
Mechanism Expert/
Novice
Content Novice
Low
Dialogue Mechanism
Familiarity
Figure
4.
Four categories of
In general, repeated
users, defined
by two independent dimensions.
exposure will increase users' familiarity level with both the contents
and mechanisms of a dialogue, but mechanism expertise tends
Regularities in the
menus can
mechanism allow
predict the
mechanism
even
that particular option before. Similarly, a user of a standard
the
number associated with
develop more rapidly.
For example, users of 2-button
transfer of learning.
for selecting an option
to
the current option by adding
if
they have never selected
menu may be
one
to the
able to predict
number associated
with the previous option, because standard menus are usually numbered sequentially.
3.2
A
Design Considerations
number of considerations govern how system
varying levels of expertise.
tests,
Where
appropriate,
we
reported elsewhere, that compared three
Virzi, et
al.
The
was
last
1992].
Two
level style choices will affect users with
cite empirical
menu
styles [Resnick
were the standard and standard with skips
a variant of 2-button menus.
It
evidence from two user
styles described above.
included a third button, to
previous item. Since users almost never pressed
13
it,
we
and Virzi 1992;
move back
to the
describe the experimental results
as if 2-button
menus were
The data from
used.
the experiments are consistent with the
design considerations but they do not provide conclusive proof.
outcome measures reported here were not
First,
some of
the
part of the original experimental design.
Second, the design considerations are stated generally for any dialogue structure, but the
experimental data comes only from applying them to menus.
DCl) A combined
action
is
simpler for users, while separate actions give them more
flexibility.
The separation of
selection from termination in a
menu
style permits users to malce
multiple selections or to change their minds about a single selection.
and then terminate can press two buttons
select an option
comes
at the
cost of increased complexity:
when
the
A
user
in succession.
two actions
are
who wants
to
This flexibility
combined, a user
need learn only a single button.
DCl)
The
Positional actions are easier to learn because they are independent of content.
selection button
another 2-button
there are
is
menu
the
same from any item
as well.
With
more mappings of actions
to
in a 2-button
menu, and from any item
the absolute selection actions in standard
in
menus,
buttons that need to be learned since the selection
button changes from item to item.
In the
second of the two experiments, subjects required several exposures
before they typed ahead numeric
styles. 2-button
the
prompt
menu
menu
users required
to a
menu
selections in the standard and standard with skips
much
less practice before they
stopped listening to
for the single select button.
This design consideration suggests an advantage for speech input over keypad input when
absolute actions are used.
mappings. That
mappings
to
is
Command names may
be easier
to
remember than button
important when absolute actions are used since there are
remember.
On
the other hand,
it
takes longer to speak a
14
command
many
than press
a button, so there
is
a tradeoff
between ease of remembering speech commands and ease
of executing button presses.
DCS) Absolute
A
actions are easier to execute than positional actions, once learned.
user can type ahead a numeric
menu
selection with a single keystroke. Positional
menu
selection requires several keystrokes or waiting until the appropriate
item plays
back.
Surprisingly, the experiments did not provide clear evidence for this. After a fair bit of
practice with the
all
three styles
that users
after
same menu
was nearly
tree
and two repetitions of complete tasks, performance with
indistinguishable. Careful analysis of the data logs indicates
were typing ahead most but not
all
of their
even more practice, users would type ahead
menu
all
selections.
We
of their selections, and then the
numeric selection styles would have better performance than the 2-button
DC4) Automatic
mechanism
transitions help
mechanism
speculate that
style.
novices, but delay their acquisition of
expertise.
A
mechanism novice can hear
in
sequence than
if
all
the parts
the caller has to learn
more quickly
if
they are automatically played
what buttons invoke
explicit transitions.
On
the
other hand, omitting the automatic transitions forces callers to learn the explicit
transitions. In the
second experiment
[Virzi, et
al.
1992], initial performance
on standard with skips menus, which include automatic
transitions, than
menus. After practice on a few menus, however, users skipped more often
style
and made selections
DCS) Automatic
was
better
on 2-button
in the 2-button
faster.
transitions, together with
'moving target' problem for
any position
sensitive actions, create a
all users.
Both the temporal and temporal with skips
styles
combine automatic
transitions with
positional selection. If a user selects just as the system automatically transitions to the
15
next
menu
option, the
action, since
it
wrong one may be
selected.
transitions to a different part
The
'next' button is also a positional
depending on the current
part. In either the
standard with skips or temporal with skips styles, suppose a user skips just as the system
automatically transitions from the
ahead
more
to the third item
first
menu
option to the second.
The system
will skip
and the user may never hear about the second. As users become
moving
expert, and select or reject items sooner, the
target
problem becomes
less
important.
DC6) When
information relevant to a user
is
preceded
in
a recording by irrelevant
information, that user will pay a time penalty.
The designer can reduce
some
to
the time penalties
imposed by recordings
users, in three ways: short recordings, explicit transitions,
an irrelevant recording
may make more
Many
may
not be so bad
if it's
that are irrelevant to
and ordering. Listening
very short. Unfortunately, novice users
errors if the descriptions of options
and available actions are incomplete.
voice mail systems handle this tradeoff by including a novice
mode with
longer
prompts and an expert mode with shorter prompts.
Menu and form
tradeoff.
As
styles that support skipping provide another
callers gain expertise, they
method
for handling this
can skip the unnecessary portions of longer
prompts, listening to just enough to cue recognition of the entire prompts.
Yet a third technique
is
to order the recordings so that irrelevant information
precedes relevant information. In general,
in
this will
never
not be possible because of differences
user interest. Designers can approach the goal, however, by putting
first
information useful to the largest number of people.
DC7) Describing
optional mechanisms degrades usability for mechanism novices
but encourages them
to
become
experts.
16
the
Optional mechanisms are those that are helpful, but not necessary. For example,
standard with skips
menu
hear
and
all
the options
key. novices
who do
gain no benefit from
style, a user
know how
The
to
may be
for teaching about the skip
use them
will
have
to listen to the
prompt, but will
to learn the skip key. If
a reasonable style choice
key that
method could be paper documentation, an
In the 2-button
it
is
still
mention the availability of the skip
however, help the user
instructions,
as identical to the standard style. This
from some other
use
about the skip button can
mention the skip key, the novice user will perceive the menu
the instructions fail to
some method
know
not
select one. If the instructions
not yet
it.
who does
in the
when
external to the interface
itself.
style
there
is
Such a
explicit training session, or transfer of learning
interface.
menu
style,
on the other hand, neither button
to hear the options
and make a selection.
A
is
optional.
A
listener
must
designer would have to be very
confident of external learning to omit instructions for necessary mechanisms.
DCS)
Questions, commands,
Users are more
and pauses encourage users
likely to wait
away.
to take action right
for additional instructions when they hear a
statement.
One
study of standard
menus [Engelbeck and Roberts 1990] measured how
users select an option immediately after hearing
it
frequently
(rather than waiting until the
end of the
menu). Subjects made fewer immediate selections with key-action wording of the menu
items ("Press 2 to do something.") than action-key wording ("To do something, press 2").
With
the former, the entire
prompt reads
as a single descriptive phrase.
interpreted the "press 2" in the latter as a separate phrase, stated as a
3.3
The
Summary
of
Users
may have
command.
User Perspective
eight design considerations above describe
various classes of users.
No one
how system
level choices will affect
dialogue style will be best for
illustrated these design considerations
by applying them
17
to a
all
applications.
few menu
styles.
We
For
example.
DC2
and
DC4
suggest that applications with few repeat callers
DC3
use temporal menus.
same menu items on each
do well with a
style that
select the
employs absolute
selection,
such as standard menus. Even subtle choices such as whether to prompt for the
may
best to
who
suggests that those applications with callers
call will
may do
serve certain types of users more than others, as suggested by
DCS. Because of
myriad choices and tradeoffs involved, however, we encourage designers
to consider
these eight points apply to their applications rather than relying on
summary
The next
The following
section further enlarges the design space for menus.
key
sicip
the
how
guidelines.
section
presents the design space for forms.
4 More
Menu
The
division of a
first
Styles
menu
recording identified a header, items, and a footer as the parts.
come from
Section 3 presented some of the styles that
recombine these
parts.
4.1
menu
how
to
This section begins with a description of some other combinations
that yield plausible styles.
particularly the
design choices about
Moreover,
items, to further
it
is
expand
useful to subdivide each of the parts,
the design space.
Combinations of Header, Items and Footer
The only
explicit transition considered in section 2
was
a "next item" action.
The only
timeout actions considered were repeat of the current item, movement to the next item.
and
restart
of the
and match them
An
menu
after the footer.
to generate
two additional menu
obvious additional explicit movement
form of the "next item" action
the
Other choices are possible.
computer advances by
3
is
relative
items.
We
some and mix
styles.
command
is
"previous item".
numeric movement: when
Menus could
18
list
A
generalized
a listener presses 3.
also include absolute
movement
;
commands
move
that
to a fixed position regardless
example, pressing 3 would move
The cautious
shown
style,
in
associated with an option to
current option,
it
to the third item.
Example
is
jump
to
it.
A user who
selected.
it
is
ABC
to the
a
the correct option.
Bank's bank-by-phone. You can jump
balances.'^ press
new
number associated with
A
the
number twice
in
variant on this style
to
number associated with
it.
any option by pressmg the number
it.
[presses
Open
number
user can press a
positional selection, so that users pressed a single selection key once
associated with
Account
to the next item,
not so cautious can press the
is
positioned on the correct item, rather than pressing the
Welcome
A
selection.
user presses the
If the
sequence, without waiting to confirm that
would use
employs timeout advance
4,
movements, and absolute
explicit absolute
of the listener's current location. For
1
interrupting j
4,
account, press 4 again.
[presses 3 J
Transfer
money between
'^
accounts, press 3 again.
[presses 3 J
Example
4:
A
The cautious
cautious
style
menu
may be
style dialogue.
especially useful to occasional users
familiar with the contents of an application without
eventually interest them. Frequently, such users
somewhere near
assumes
the
may remember
end of the menu, without remembering
that it's easier to prevent errors than to recover
guess the number. The style
to skip
memorizing
through
all
is
It is
all
the options that
may
that the desired option is
exact number. This style
from them:
also reasonably effective for
the options quickly.
its
who become somewhat
it
makes
it
safe to
mechanism experts who want
not quite as effective, however, as having a
single skip-ahead key because the user has to press one, then two, then three, rather than
pressing the same key repeatedly.
Remembering
the
number
for the next position
is
an
added cognitive burden, as well as being mechanically more cumbersome. Finally, the
19
style functions similarly to standard
the current option just after the
Timeouts can be used
menus
number
for
it
escape from the
to
for
is
complete novices, as long as they select
announced.
menu
or to
make
selections, instead of or in
addition to advancing to the next item. For example, call processing applications often
use a variant of standard
human
menus where
a timeout at the
operator, rather than repeating the menu.
then escape
if
the user
still
Some even
repeat the
to a
menu once and
has not selected an option.
menus (Example
Rejection
end of the menu escapes
5) are
one
style
where a timeout
selects the current option. In
this style, users press a button to reject the current option or wait to select
it.
This style
is
a counterpart of the temporal style described above where users press a button to accept
the current option or wait to hear the next. If
(see
to select
DC6
Welcome
If
you'd
is
desirable to use only one button,
menus have some advantages over temporal menus,
rejection
wants
it
an option
late in the
menu and can
particularly
when
a user
quickJy reject some of the earlier ones
above).
XYZBank's bank-by-phone.
to
like
account balances, please wait. Otherwise, press
#.
[presses #/
Transfers between accounts. Wait or press # to reject.
[presses #/
Mortgage ^
Open
a
rates.
#
[presses
#.
new
to reject.
interrupting!
account. # to reject.
[waits, causing this option to be selected!
Example
5:
A
sample rejection menu dialogue.
4.2 Subdivision of
It is
Menu Items
useful to further subdivide the header, item, and footer recordings, both to clearly
identify the kinds of information they
convey and
20
to allow for explicit
movements
betu-een s_rr_r_^.
•'•
t
—
.-'S2ie this subdi\"ision
and recombination process for menu
Items.
A ~r- -
-
contain three kinds of information: a description of the contents, an
indicator of whether
it is
already selected (when multiple selections are allowed) and
informaticr. _: _-: available actions. T:.e ;^.=ction indicator could be a tone, a
as "selected." or a
change of voice for the option description
word such
male instead of
(e.g..
female i. Just as the header, items, and footer can be recombined in more than one
iznr. w.::ereni si>ies. so the subparts of
move through
Typically. l~cc-L5
menu
is
the subparts
st\les that pro\ide an explicit skip
generally best to put Lhe action prompts
to
items can be recombined in different ways.
m a particular order,
~e-.- :tem as a single part. In that case, the order of the subparts
by DC6. In
way
key for movement
at the
end. where
giving the effect of a
is critical,
to the ne.xt
mechanism
as suggested
menu
item,
it
e.xperts will not
be bothered by them. For standard menus, too, most researchers agree that the selection
prompts should follow the option prompts
1990; Halstead-Nussloch 1989]. Thus,
:e— :-:—. pts in
.
There
is
all
menu
it
("'for
X. press 3") [Engelbeck and Roberts
appears that the action prompts should follow
st\les.
a tradeoff in whether to inclu(ie action prompts for actions that are helpful but
not necessarv' in interacting with the menu. For example, during pilot testing for the
experiment cited above, we tned mree variants of the standard with skips
told users in the header that they could press
#
in
any of the menu items. Some of our
#
to skip ahead.
One
pilot subjects pressed
stvle. .Ail three
variant did not mention
#
to skip
menu
headers,
but never guessed that they also could skip through the options. The second variant
ned pound after each item. Subjects were very slow
them
in the
header and the
variant for the final
first
initially.
The
item in each menu, but not thereafter.
smdy because
it
third variant told
We
chose
produced the best overall performance
lesL
21
m
this last
the pilot
Menus can
also include explicit actions for
possibility
is
moving among
the subparts of an item.
provide a "help" action that moves from the option description to the
to
beginning of the prompts for available actions. Consider the fast standard
Example
illustrated in
6.
Like standard menus,
Each item consists of a
selection.
One
uses timeouts to advance and numeric
it
terse option
st>-le.
descnption and a prompt
to press the
number
associated with the option. Unlike standard menus, however, the prompt for the
number
is
menu
not played unless the user presses the help key (0). Thus, while listening to the
items, the
menu sounds
like a
temporal menu, since
it
includes only option
descriptions, not selection prompts. For users unfamiliar with the
menu
contents, this
allows them to hear the options more quickly than with the standard
st>le.
Once
becomes
familiar with the
menu
contents, however,
it
is still
the user
possible to type ahead a
numeric selection, which would not be possible with temporal menus.
Although the
fast standard st>le
drawback:
assumes users
it
will
may be worth
know
to press
exploring further,
when
it
has one major
they want to find the
number
associated with a particular option. The header can include a prompt that mentions
may
but that
not be
to find out
users to press
the
numenc
Welcome
enough
for first-time users. If the
how
to select, that
menu
prompt
this.
items include a prompt telling
will take
more time
to recite than
selection crompt.
to the
.\BC Bank's bank-by -phone. Press
when you hear
the
opuon
that interests you.
Account balances:
Transfer
money between
Mortgage
Open
a
accounts;
rates:
new account
^:
[presses 0]
Press 4 to select this option. In the future, you can press 4 to select this option an> time
[presses 4]
Example
6.
The
fast
standard st\le.
10
dunng
the
menu.
'^
4.3
Prompt Wording
Variations
even
wording of prompts can give two
in
they are identical on
if
all
the
styles a very different
sound and
feel,
system design dimensions. For example, a variation
of the standard with skips style was designed for obsessive-compulsive psychiatric
patients
who could
prompt
for
not tolerate ambiguity
each menu item
immediately following
al.
1993].
It
included both a short
"checking account balance") and
(e.g.,
(e.g.,
[Sorce, et
longer one
a
"the account balance for checking account
number
1042030776.")
The yes-no
illustrated
style,
differentiating factor
is
that the
in
Example
7.
is
a variant of 2-button
yes-no style uses interrogative prompts phrased as yes-no
questions whereas the 2-button style uses descriptive prompts. The
menu
uses a
initial
interrogation phrase
Welcome
to the
Do you want
first
item
in a
yes-no
sentence question and subsequent items use fragments that omit the
full
ABC
to hear
menus. The
("Do you
want...").
Bank's bank-by-phone.
your account balances?
1
yes, 2 no;
[presses 2]
transfer
money between
accounts?,
1
yes, 2 no;'^
(presses 2]
mortgage rates?
1
'^
yes,
2 no;
[presses 2, interrupting prompt]
open a new
'^
account?
[presses
Example
7.
1.
yes. 2 no;
The yes-no
Another variation
in a
1
interrupting prompt]
2-button
is
menu
to
style, a variant
on 2-button menus.
have prompts for actions draw on
style variant, the
spatial analogies.
For example,
keys on the telephone keypad can be used as cursor
23
keys (4
The
1985].
down
left,
6 right, 2 up, 8 down) [Robens and Engelbeck 1989; Rosson and Melien
action prompts could be,
to select the current option;
4.4 Lists: a Special
Many
Case
of
menu
without making a selection".
Menus
as dialogue
lists
move through
as degenerate cases of
lists
right for the next option; left for the previous;
to exit this
audio applications also include
mail applications allow a user to
view
up
"Go
menus
a
list
that allow
components. For example, voice
of messages
movement
in a
mailbox.
We
can
but not selection. For
example, a conventional cassette-based answering machine provides the analog of
temporal menus:
plays the messages one after the other, using timeouts to advance
it
between messages. All the menu
equally applicable to
One
list
style is
commands.
We
list
style variations (except those relatmg to selection) are
styles.
worth analyzing because
call
it
includes several unusual
it
the radio-scanner style because
radio scanners found in
it
makes an
many automobiles [Kondziela
station to station, playing a
few seconds of each
1990].
movement
explicit analogy to the
The
radio scans from
until the user presses a button to stop
scanning. Similarly, the radio-scanner style advances from item to item via timeouts,
playing just a headline of each item. In addition to timeout movements, # and
explicit
move forward and move back commands. ## moves ahead by
back by
five items. This restricted use of relative
numeric movement
five
*
are
and ** moves
all
of the
numeric keys available for absolute movement. Rather than pre-assigning numbers
to the
items in the
list,
still
leaves
each user can assign numbers to favorite items. The user does so by
pressing the numbered button twice while listening to that item. Thereafter, pressing that
number once
The
initiates
radio scanner
list
an absolute movement
to that item.
style also includes an explicit action for
Each item consists of a headline separate from
24
movement within an
the rest of the item.
To
get
item.
from the
headline to the rest of the item, a user presses a button (0 in this case). This
similar to the explicit help variation used in the fast standard
user presses
the
Summary
4.4
computer plays
of
menu
style, but
here
is
quite
when a
the rest of the item.
Audio Menus
The primary design space
for
menus comes from
a division of the recording into a header,
items, and footer, and the division of the select action into
marking and termination.
We
presented several styles in this space and suggested applications to which they might be
especially well suited.
the subdivision
menu item
We
expanded
that
design space through a recursive application of
and recombination framework
there
were opportunities
illustrated. Finally, variations in
same system design choices
to the
menu
items.
Even within
a single
for explicit transitions, as the fast standard style
prompt wording can give two implementations of the
a very different sound and feel.
We turn now
to
audio forms.
5 Forms
Forms guide people through
the process of entering several related pieces of information.
This section begins with three sample form styles, to
diversity.
illustrate
are a header, a footer,
styles. In this case, the
one entry blank for each piece of information
optionally a review node associated with each entry blank.
values from entry blanks.
5.1
the potential
Then, the analytic framework of dialogue mechanism decomposition and
recombination generates a design space for form
form
some of
We
The
to
recording parts
be entered, and
actions insert and
remove
then recursively subdivide the entry blanks: two additional
styles illustrate points in the
expanded design space.
Three Form Styles
The
first
and most easily understood telephone form was part of the PhoneSlave
[Schmandt and Arons 1984], which took phone messages when
from
his desk.
It
used a conversational
style.
25
its
'master'
The system asked each
was away
caller a series of
]
questions ("Who's calling please", "What
he reach you?",
etc.)
is this in
After playing a question,
it
reference to?", "At what
number can
recorded whatever the caller said, until a
long pause was detected, then went on to the next question. Example 8 illustrates this
style for a classified advertising application
where the user enters information about a car
for sale.
What
kind ot car are you selling?
Entry
["Cadillac"!
Please enter the year? For example, enter eight-six for a 1986 model.
[presses
What
color
is
9.
then IJ
the car?
[says "Gray... well, more bluish-gray
'
What
Entry blank
kind of car are you selling?
o
[••Cadillac'"]
You
said. "Cadillac" (plays
press #. If not. press
back recording).
r
If that's right,
1
an )a
Review node
*.
1
[Presses #]
Please enter the year? For example, enter eight-six for a 1986
model.
[presses 9, then
A
1991 model.
I
Entry blank 2
1
]
If that's right, press #. If not. press *.
o
r
[Presses #]
What
color
is
Review node 2
the car?
more bluish-gray "]
more bluish-gray If that's
[says "Gray... well,
You
said, "Gray... well,
[Presses
What color
is
*.
".
interrupting prompt]
said, "Bluish gray.
"
If that's
'^
[Presses #. interrupting prompt]
Enter your phone number.
[presses 2-2-2-9-9-9-9]
222-9999.
If that's right,
^
[Presses #, interrupting prompt]
Example
9.
The
H
Entry blank 3
the car?
[says, "Bluish gray"]
You
'^
careful
form
style.
y
Brand.
#.
To
record, press
1
End recording by pressing
.
For the next entry blank, press
[Presses
Model
year.
To
Entry blank
says "Cadillac", presses #]
i.
enter a value, press
previous entry blank, press
7.
1
To review
.
the
For the next entry
jm
1]
Enter two digits. For example, enter eight-six for a
1986 model.
Entry blank 2
[Presses 9, then
Color.
To
1
]
begin recording, press
[Presses
QD
or
blank, press 9.
[Presses
1
9.
1.
'^
says "Gray, well, more bluish-
1,
gray]
DP
o*-
Phone number. To enter
a value, press
the previous entry blank, press 7. ^
To review
I.
^
[Presses 7]
more
Color. "Gray, well,
recording, press
I.
[Presses
1,
To
replace this
Entry blank 3
says, "Bluish-gray"]
Phone number. To enter
[presses
bluish-gray".
^
a value, press
^
1.
at
or
any time.
^
[presses 2-2-2-9-9-9-9]
That's the end of the form.
this
ad and would
like to
If
save
you're satisfied with
it,
press
'^
3.
an
w
1]
Enter your seven-digit phone number
m
[presses 3]
Figure
Example
10.
These three
The
styles only hint at a larger design space.
populations, conversational forms
incorrect entry are high, however,
either automatically or
user initiation
is
suitable choice.
the
7.
user-controlled form style.
upon user
may
For some applications and user
be appropriate.
When
the consequences of
some method of allowing review should be provided,
request.
If
some of
the entry blanks are optional, or
desirable, the user controlled style or variations
Sometimes, none of these three styles
consequences of incorrect entries are very high,
it
will
may
initiation of value entry, as in the user-controlled style, but
value, as in the careful style.
28
on
be quite
it
may be
right.
the
most
For example,
if
be appropriate to have explicit
still
automatically review each
5.2
We
Design Dimensions
can gain more insight into
tiie
style variations
associated with recombining dialogue parts.
into a header, a footer, entry blanks,
example, the
first
entry blank in
by exploring the design dimensions
The recordings
in a
form divide naturally
and review nodes associated with entry blanks. For
Example 9
(the careful style) stated,
you selling?" After the user recorded, "Cadillac", the review node
are
'Cadillac'. If that's right, press #. If not, press *."
To conserve
"What kind of
stated,
"You
car
said,
space, the examples
omitted the headers and, where possible, the footers; they function analogously to their
The primitive
counterparts in menus.
actions in a form are insertion and deletion of
values from entry blanks. The design choices include which actions are available for
movement, how value changing
actions are bundled, and whether they are initiated
explicitly or with timeouts.
5.2.1
The
Movement Actions
possible transitions for
movement among items
all
in
the entry blanks are analogous to those for
a menu. Since users will typically enter information in most or
vement forward and back by one entry blank are appropriate
of the entry blanks, n
movement commands.
movement among
Fr
i
the review nodes, users can return to the associated entry
blank or transition to the next one. The backward movement action
influences users' ability to review, especially
controlled style,
it
allows review of values
if
is
one factor
that
there are no review nodes. In the user-
at the user's discretion,
by explicitly moving
backwards.
All the form styles in this paper include a subset of these
applications, however,
accommodate
may
benefit from additional
movement
actions.
mechanisms. For example,
selective review of long forms, absolute or relative numeric
be included, or actions to
move
to the next
empty or
29
Some
to
jumps might
the next already filled entry blank.
An
action to
blanks.
The
jump
inclusion of
complexity for users.
transitions,
If,
more
be useful after reviewing the contents of a few entry
explicit transitions does not necessarily increase the
however, the system includes prompts for the additional
mechanism novices
5.2.2 Action
may
to the footer
will
pay a time penalty (DC7).
Combinations
Ail form styles need
some way
to
add a new value
to an entry blank.
Some
may
styles
include a delete action as well, to remove a value from an entry blank.
Several composite value change operations
command
erases not just one but
delete followed by an insert.
and then
initiate
flexibility
The
when
A
insenion of a
all
one.
all
command would
As described
actions are separated and simplicity
insertion action
is
included. For example, a delete-all
the values in an entry blank.
replace
new
may be
in
replace action
there
is
a tradeoff
actions are combined.
often bundled with a transition to the review node or the next entry
subroutme may allow the user
from a menu.
bundled
transition.
a
between
blank. After initiation of value entry the form passes control to a subroutine.
select
is
delete all the current values
DCl,
when
A
When
to record, enter a
sequence of touch-tones
the value entry subroutine returns, the
The conversational and user controlled
(e.g.,
The
a date) or
form follows the
styles transition to the next
entry blank while the careful style transitions to a review node.
From
the review node, a
user can erase the value and return to the current entry blank, or go on to the next entry
blank.
Of course,
it is
not necessary to bundle the insertion action with a
movement
in this
way.
After returning from the value entry subroutine the system could replay the contents of
the current entry blank. This
may
be appropriate either as an alternative mechanism for
reviewing the contents of the current entry blank, or
values to the entry blank.
30
to
encourage the addition of several
Other composite actions may also be included. For example, an undo
moves
to the previous entry blank,
review node, the undo
and then deletes a value
command would
command
there. Similarly,
return to the current entry blank
first
from
a
and erase the
value just entered.
5.2.3 Positional vs. Absolute Initiation
It is
theoretically possible for an absolute
from anywhere
in the
mechanism
to initiate value entry or
removal:
form, a user could initiate addition of a value to any of the entry
blanks. All the form styles in this paper, however, utilize positional initiation: insertion
and deletion actions apply
to the current entry blank.
5.2.4 Effect of timeouts
Entry of a new value can be initiated with an explicit action, or with a timeout. This
choice
may have
the single largest effect
on the
feel of the form. Explicit initiation of
value entry gives users control over the pace of the interaction, allowing them to gather
their thoughts before entering information.
On
the other hand, timeout initiation can
make
the dialogue flow naturally for novice users (DC4).
When
value insertion
by default, even
is
the only action available
the user does not wait for the timeout at the
if
example,
in
interprets
any user input as a date.
From
from an entry blank,
it
can be
initiated
end of the entry blank. For
an entry blank that expects input of data by touch-tones, the careful style
a review node, timeouts can initiate any of the possible actions. For example, in the
careful style, callers
blank, or confirm
it
repeats the prompt.
must either
and go on
An
explicitly erase the value
and return
to the next entry blank. If the user
alternative version
would
31
to the current entry
does neither, the system
treat silence as assent
(timeout
moves
ahead
to the
next entry blank) or dissent (timeout erases the value and returns to the
current entry blank).
5.3 Entry
As
in
Blank Subdivision
menus,
it
is
useful to apply the subdivision idea recursively, in this case to entr\'
blanks. This highlights the types of information that an entry blank normally includes and
opens additional design choices for recombination of the
making
the insertion and deletion
commands
parts. In particular,
we consider
sensitive to the current position within the
entry blank.
Three kinds of information can appear
in
an entry blank: a description of the desired
values, current values (e.g., recordings), and prompts for actions.
desired values can either be descriptive
("Record the color of your
initiation
informal
car").
DCS
(e.g.,
"The color of your
The descriptions of
car") or
commanding
suggests that the choice of timeouts or explicit
of actions interacts with the choice of wording styles for these descriptions. In
we found command
tests,
of value entry.
When
They proceeded
statements to be less effective with explicit initiation
entering dates, users often forgot to press
1
to initiate data entry.
which
to enter several touch-tones (e.g., 0-7-3-1 for July 31),
the
system
interpreted as explicit user actions rather than as entry of a date.
Thus
far, the
discussion has assumed that value change
delete have the
same
effect
from anywhere
in
commands such
an entry blank.
When
as insert
and
entry blanks can
include multiple values, position sensitive actions can be helpful. Consider an entry blank
that has several dates,
always remove the
each entered by touch-tones.
last date.
A
An
absolute deletion
positional deletion action could
remove
command would
the date currently
being played. Likewise, the user could insert an additional date just before the current
one.
32
The same
idea applies to entry blanks that contain recorded voice. Positional insertion
and deletion are especially useful
when
a user inserts a
two segments and
action could
new recording
inserts the
remove
in dictation applications.
new
computer
recording
in the
splits the
would require
voice portion to delete. Even
in fairly
insertion,
recording then playing into
middle. Similarly, a positional deletion
segment currently playing back.
the voice
positional deletion action
the
With positional
the user to
mark
An even more complex
the beginning
and end of the
complex dictation applications, however,
the
simpler mechanism of deleting the entire voice segment currently playing might work
quite well.
Two
variants of the user-controlled style illustrate the design possibilities for entry
blanks.
The event calendar
style
new
[Resnick 1993] allows the general public to add
event announcements to a public bulletin board. Callers
for a headline, date, time, location, sponsor, contact
fill
out forms with entry blanks
phone number, and
details.
The
style
has evolved over the more than two and a half years that the application has been used by
the general public.
The event calendar allows multiple
appending additional thoughts
filling in a
deletes
values. This
is
frequently useful for dates and for
in the details entry blank. In
form for a classified
ad, enters
1
1,
the user,
who
is
two telephone numbers. The delete action
the voice in the entry blank, not just the last
all
Example
segment recorded, because
informal tests indicated that was too confusing for novices. The delete action does,
however, delete only the
last
date
Brand. The kind of car you are selling.
ready, press
1.
End recording by
[Presses
Year.
1,
if
more than one has been entered
You can
an entrv blank.
gather your thoughts before starting to record.
When
you're
pressing #. For the next entry blank, press 9.
says "Cadillac", presses #]
The model year of
the car.
To
enter a value, press
1
.
To review
the next entry blank, press 9.
[Presses
in
1]
Enter two digits. For example, enter eight-six for a 1986 model.
33
the previous entry blank, press 7. For
[Presses 9, then
Color.
The color of your
1
]
car.
To begin
recording, press
^
I
.
[Presses 1, says "Gray, well, more bluish-gray]
Phone number. The number people should call if they want
review the previous entry blank, press
7.
to
buy your
car.
To
enter a value, press
1.
To
enter a value, press
1
To
'^
[Presses 7]
Color. "Gray, well,
[Presses
1.
more
bluish-gray".
To
I.
'^
says. "Bluish-gray"]
Phone number. The number people should
[presses
replace this recording, press
call if they
want
to
buy your
car.
.
^
1]
Enter your seven-digit phone number
at
any time.
[presses 2-2-2-9-9-9-9]
That's the end of the form. ^
[presses 7]
Phone number. 222-9999. To enter an additional
[press
value, press
1.
'^
1]
Enter your seven-digit phone number
at
any time.
[presses 3-3-3-8-8-8-8]
Example
11.
entry blanks
A
sample dialogue with the event calendar
initially
style.
Note
that in this style,
have long descriptions. They are omitted once a user has entered a
value.
The
dictation style
is
another variant of the user-controlled style.
It
permits multiple
values in entry blanks and employs positional insertion and deletion actions. The style
might be well suited
to the dictation of entries for a patient chart in a hospital.
chart provides a single place
where
all
doctors and nurses
who
The
patient
minister to a patient
make
notes intended for the use of the other caretakers (and for the courts in the case of
malpractice claims). In most hospitals, patient chart entries are free-form, either written
by hand or dictated and then transcribed. There
is
no pre-defined structure, but each
hospital has conventional genres for what information to include in
specific kinds of entries.
Some
forms would have symbolic
of these conventions could be
fields,
such as the doctor's
quantities, as well as unstructured fields to
embedded
id
what order
in forms.
for
Such
and drug prescription
accommodate spoken
notes. Doctors and
nurses could add user-defined structure by recording several separate segments in each
entry blank.
34
Brand. The kind of car you are selling. Record,
[Presses
1.
1
;
Next entry blank.
9.
says "Cadillac", presses #]
Brand. "Cadillac". Record.
1;
Erase. 2; previous entry blank, 7; next entry blank, 9.
[Presses 9]
The
Year.
two
last
[Presses
Enter two
digits of the
1.
model year of
the car. Enter value.
1
;
Next entry
'^„,
interrupting prompt|
For example, enter eight-six for a 1986 model.
digits.
[Presses 9. then
1]
Year. 1991. Enter value, ^„.
[Presses 9, interrupting prompt[
Color.
The color of your
[Presses
1,
car.
'^
says "Gray, well, more bluish-gray]
Color. "Gray, well, ^ more bluish-gray"...
[Presses
1.
interrupting, and says "Actually, it's", then presses #]
Color. "Gray, well, actually,
it's
more bluish-gray"
'^
[Presses 9]
...and the
dialoeue continues.
Example
5.4
12.
The
dictation
form
style.
Form Styles Summary
Form design
actions and
described.
vectors summarize the design choices about
how
The
movement and value change
they are combined. Figure 8 shows the design vectors for the five styles
first
three
show
the effect of design choices about
entry blanks and actions that change values in entry blanks.
The
last
movements between
two
illustrate
design
choices for actions within individual entry blanks. For the sake of simplicity, the design
vectors omit information about which subparts to include in each main part and
word prompts.
35
how
to
Action
generate
we had
new
styles.
We
had not considered
all
of the styles presented
paper
in this
until
constructed the dimensions of the design space.
With so many
styles to
choose from, and the possibility of generating custom
styles, the
designer can tailor the choice to the needs of particular applications, just as screen-based
menu
interface designers choose
styles (e.g., pull
down, pop
particular needs. For example, if users will frequently
it
will be
worth
in
menus.
If
options,
options
selection
may
be best.
If
frequent users will
some
commands
memorize
for
movement among
the contents of
menus, numeric
entry blanks in a form are optional or if users will need
time to gather their thoughts before recording, a form style with explicit
initiate
value entry
may
match
browse among large collections of
their initial effort to learn the
some
up, radio buttons) to
be appropriate. Rather than relying on these
however, we encourage designers
to describe their
commands
summary
to
guidelines,
expected users and apply the design
considerations of section 3 to a range of possible styles.
While the design spaces are useful
dialogues, they have
and personal
tasks.
much
to researchers
broader implications. Audio interactions with workstations
digital assistants will
be useful whenever the users' eyes are busy with other
Moreover, much of the analysis
temporal
in nature,
even
if
and designers of telephone-based
it
is
information can be presented
in this article carries
not audio. In general,
at
one time, then
all
if
initiate
how
users can navigate
among
is
only a small part of the relevant
of the fundamental audio design
choices will be relevant: what information to include in
blanks,
over to any interface that
menu
options and form entry
the pieces of information,
and which actions
to
with timeouts versus explicit commands.
Consider, for example, the
new
generation of ADSI-compatible display phones [Bellcore
1992]. These phones have displays that are 20 characters wide and up to eight lines long.
There
is
a signaling protocol that allows these
during the course of normal voice
calls.
One
phones
to receive
natural idea
37
is
to
160 character batches
augment
interactive voice
response services with a presentation of menus on the screen.
on each
menu
line,
it
may
be possible to display the entire
menu
at
one option
If
is
displayed
once, but 20 characters per
option leads to cryptic descriptions of the options. Another possibility would be to
display one option at a time, using the
option. Since only one option
variables apply, including
at
how
full
screen space to give a clearer description of the
a time will be displayed,
users
move through
many
of the audio
make
the options and
menu
style
selections.
Similar design considerations apply to other small-screen devices, such as the screen
employed
telephony services [Karis,
in in-flight
system [Fitzmaurice,
et al. 1993].
necessary information
presented in this paper
The temporal
1993] or a palmtop virtual reality
These authors point out the
in a limited
may
et al.
difficulty in presenting the
screen space. Analogs of the
menu and form
styles
apply to such devices.
presentation of audio creates interesting design challenges. Exploration of
techniques that give users control over the time dimension are just beginning. The
essence of that exploration
natural
ways
this article
is
to
break information chunks into ever smaller parts and find
for users to control
which
have applied that principle
and forms into component
possibilities.
It
may
parts,
part will be presented next.
to
The design spaces
in
both menus and forms. The division of menus
and those parts into even smaller
be that the best styles have not been invented
38
parts,
yet.
opens up new
References
Arons. B. Techniques and Applications of Time-Compressed Speech. In Proceedings of
'92: Conference of the American Voice Input/Output Society (1992, Minneapolis,
AVIOS
MN).
Arons. B. SpeechSkimmer: Interactively Skimming Recorded Speech. In Proceedings of
UIST '93: Symposium on User Interface Software and Technology (1993, New York:
ACM),
pp. '187-196.
Bellcore Generic Requirements for an
Interface of Analog Display Services
(February, 1992;
.
SPCS to Customer Premises Equipment Data
Technical Advisory
#TA-NWT-001273,
Card. S. User Pe .eptual Machanisms in the Search of Computer Command Menus. In
Proceedings of CHI '82: Human Factors in Computing Svstems (1982, New York:
ACM), pp. 190-196.
Your Fingers Do the Spelling: Disambiguating Words Spelled with the
Keypad.
Journal of the American Voice I/O Society, 9, March (1991), pp. 57Telephone
Davis. J.R. Let
66.
Degen, L., Mander. R. and Salomon, G. Working with Audio: Integrating Personal Tape
Recorders and Desktop Computers. In Proceedings of CHI '92 Conference on Human
Factors in Computing Systems (1992, Monterey: ACM), pp. 413-418.
Jr. and Gattuso, N., Jr. Alphabetic input on a Telephone
Keypad. In Proceedings of the Human Factors Society-- 34th annual meeting. (1990,
Santa Monica, CA: Human Factors Society).
Detweiler, M., Schumacher, R.,
Engelbeck. G. and Roberts, T. The Effects of Several Voice-Menu Characteristics on
Selection Performance
US West Advanced Technologies #Technical Report
ST0401,(, 1990).
Menu
Fast, L.
.
and Ballantme. R. Dialing a name: Alphabetic Entry Through a Telephone
Keypad. SIGCHI Bulletin, 20, 2 (1988), pp. 34.
Fitzmaurice. G.W., Zhai, S. and Chignell, M.H. Virtual Reality for Palmtop Computers.
Transactions on Information Systems, 1 1, 3 (1993), pp. 197-218.
ACM
Gould, J.D. and Boies, S.J. Human Factors Challenges in Creating a Pricipal Support
Office System: The Speech Filing System Approach. ACM Transactions on Office
Information Systems, 1, 3 (1983), pp. 273-298.
Halstead-Nussloch. R. The Design of Phone-Ba.sed Interfaces for Consumers. In
Proceedings of CHI '89 Conference on Human Factors m Computing Systems (1989,
Austin,
TX: ACM),
pp. 347-352.
Hindus, D. and Schmandt, C. Ubiquitous Audio: Capturing Spontaneous Collaboration.
1993).
Special Issue on CSCW,
ACM Transactions on Information Systems,
,
39
(
Karis, D.. Norton. S. and Dobroth, K. User-interface Design for an In-flight
Telecommunications Service. In Proceedings of Human Factors in Telecommunications
(1993. Heidleberg, Germany: R.v. Decker's Verlag), pp. 51-62.
Kate, Y. and Hosoya. K. Message Browsing Facility for Voice Bulletin Board Service.
Proceedings of Human Factors in Telecommunications 1993, Heidelberg, Germanv
R.v. Decker's Verlag), pp. 321-330.
In
(
Kiger, J.L. The Depth/Breadth Trade-off in the Design of Menu-Driven User Interfaces.
International Journal of Man-Machine Studies, 20. 2 (1984), pp. 201-213.
Kondziela, J.M. Radio-Analogue Method of Audio Presentation
Services
US Patent #4,899,369, (Feb. 6, 1990).
and Control for Audiotex
.
Landauer, T.K. and Nachbar, D.W. Selection from Alphabetic and Numeric Menu Trees
Using a Touch Screen: Breadth, Depth, and Width. In Proceedings of CHI '85 (1985,
New York: ACM), pp. 73-78.
Lee, E. and MacGreeor, J. Minimizing User Search
Factors. 27,^(1985), pp. 157-162.
Time
in
Menu
Retrieval Systems.
Human
MacGregor,
a Decision
Maries,
J.,
Lee, E. and Lam, N. Optimizing the Structure of Database Menu Indexes:
Search. Human Factors, 28, 4 (1986), pp. 387-399.
Model of Menu
M. How Do You Enter "D'Anzi-Quist Using
"
Proceedings of the
Human
CA: Human Factors
Keypad?
a Telephone
Factors Society— 34th annual meeting
(
In
1990, Santa Monica,
Society).
Miller, D.P. The Depth/Breadth Tradeoff in Hierarchical Computer Menus. In
Proceedings of Human Factors Society 25th annual meeting (1981. Santa Monica, CA:
Human
Factors Society), pp. 296-300.
Norman. K.L. The Psychology of Menu Selection: Designing Cognitive Control of the
Human/Computer Interface. Ablex, Norwood, NJ, 1991.
Faap, K.R. and Roske-Hofstrand, R.J. The Optimal
Factors, 28, 4 ( 1986), pp. 377-385.
Number
of
Menu Options
per Panel.
Human
Pierce, B.J., Parkinson, S.R. and Sisson, N. Effects of Semantic Similarity, Omission
Probability and Number of Alternatives in Computer Menu Search. International
Journal of Man-Machine Studies, 37, (1992), pp. 653-677.
Resnick. P. Phone-Based CSCW: Tools and Trials.
Systems, October 1993, (1993).
ACM Transactions on Information
P. and Virzi, R.A. Skip and Scan: Cleaning Up Telephone Interfaces. In
Proceedings of CHI 92: Conference on Human Factors in Computing Systems (1992,
Monterey, CA: ACM), pp. 419-426.
Resnick,
Roberts, T.L. and Engelbeck, G. The Effects of Device Technology on the Usability of
Advanced Telephone Functions. In Proceedings of CHI '89 Conference on Human
Factors in Computing Systems (1989, New York: ACM), pp. 331-337.
40
Rosson. M.B. and Mellen, N.M. Behavioral Issues in Speech-Based Remote Information
Retrieval. In Proceedings ofAVIOS ( 1985, San Francisco, CA).
Schmandt, C. Voice Communication with Computers.
In press.
Schmandt, C. and Arons. B. A Conversational Telephone Messaging System. IEEE
Transactions on Consumer Electronics, CE-30, 3 (1984), pp. 21-24.
J.F., Baer, L.. Henriques, A. and Brown-Beasley, M. Applying Interactive Voice
Response Technology in Psychiatric Practice. In Proceedings of Human Factors in
Telecommunications (1993, Heidelberg, Germany: R.v. Decker's Verlag), pp. 397-407.
Sorce,
A
Stifelman. L.J., Arons, B., Schmandt, C. and Huiteen, E.A. VoiceNotes:
Speech
Interface for a Hand-Held Voice Notetaker. In Proceedings oflNTERCHI '93 Conference
on Human Factors in Computing Systems (1993, New York: ACM), pp. 179-186.
Virzi, R.A., Resnick, P. and Ottens, D. Skip and Scan Telephone Menus: User
Performance as a Function of Experience. In Proceedings of Proceedings of the Human
Factors Society (1992, Santa Monica, CA: Human Factors Society), pp. 211-215.
41
MIT LIBRARIES DUPl
^illHiillini'iii
nil!
3
TDSO OOflSb^ai
T
Date Due
Lib-26-67